An empirical sequence to extract fetal electrocardiogram using the Kernel and wavelet optimization

Abstract

Fetal electrocardiogram (ECG) monitoring from the mother’s abdomen during pregnancy provides electrophysiological statistics of fetus heart. To separate fetal ECG from mother ECG is a signal separation technique that has the potential to give blind source information. In this paper, an empirical sequence of Kernel Independent Component Analysis (KICA) and Dual-Tree Complex Wavelet Transform (DTCWT) is implemented for efficient ECGs separation and post enhancement of extracted fetal ECG. The kernel approach provides a non-linear transformation and works in building contrast function based on mutual information. Further to remove unwanted disturbances, shift-invariant DTCWT is used to reconstruct filtered extracted fetal ECG. The proposed work is validated on Abdominal and Direct Fetal ECG Database (ADFECGDB) from the physionet. The proposed method also compares the accuracy and running time by implementing the FastICA method on the same dataset simultaneously.